How confidence works
The confidence card on every venue page is built from real diner outcomes — not a star average. Here is exactly how the evidence is weighed, in the order it matters.
What is the confidence signal?
The confidence signal is how feefrae weighs real diner outcomes for one venue, scoped to your allergy and severity. It reflects who reported, what happened, how recently, whether the venue’s own claims held up, and which way the evidence is trending. It decides which venues you see first — but it is never shown as a single score, a star, or a safety verdict.
- 1
Who reported — matched to you
Every report carries the reporter’s allergy and severity, frozen as at the visit. A coeliac sees evidence from coeliac diners; someone with a severe nut allergy sees it from people managing the same. A reassuring report from a mild intolerance never counts as reassurance for anaphylaxis — severity is evidentiary weight, not just a filter.
- 2
What happened — the outcome
The load-bearing field isn’t a star, it’s the outcome: no reaction, a reaction, a near-miss (“something felt off”), or unsure. A near-miss lowers the signal as an early warning, before anyone is harmed. One reaction among many clean visits is shown in proportion — never as a verdict.
- 3
How recently — freshness
Allergy trust decays: staff, chefs, suppliers and menus all turn over. A glowing report from three years ago does not carry the weight of one from last week. Evidence is labelled fresh, ageing or stale, and old evidence quietly fades on its own.
- 4
Does the claim hold up
When a venue states something — “dedicated gluten-free fryer”, “separate prep” — diners who ate there say whether they actually saw it. We show the agreement (or the gap) and grade how much evidence exists: strong, limited, or mixed. That grades the evidence, never whether the venue is safe.
- 5
Which way it’s moving — the trend
Diners trust a direction, not a snapshot. We combine recent outcomes and freshness into a trend — improving, stable or declining — plus any recent change. A declining trend or a fresh change report is the most decision-relevant thing we can show you.
8 diners · 19 visits · 6 repeat
- 17 reported no reaction · 1 near-miss · 1 unsure
- 6 reported a dedicated fryer · most recent visit 12 days ago
- Confidence improving recent change: allergen matrix updated
Read it out: eight coeliac diners, six of whom went back; almost all with no reaction, one early-warning near-miss kept visible; recent, and trending up after the venue refreshed its matrix. That’s a richer, more honest picture than “4.6 stars” — and it still isn’t a promise. Always confirm with the venue, and carry your medication.
Frequently asked questions
What is feefrae’s confidence signal?
It is how feefrae weighs real diner outcomes for one venue, scoped to your allergy and severity — who reported, what happened, how recently, whether the venue’s claims held up, and which way it is trending. It orders which venues you see first, but it is never displayed as a number or a star.
Why don’t you show a single trust score or star rating?
A branded number gets read as a safety verdict — exactly what feefrae refuses to issue. The score exists internally to order results, but the card shows honest counts (visits across how many diners, what they reported, how recent) rather than collapsing that into one figure.
How does allergy severity change the weight of the evidence?
Severity is evidentiary weight, not just a filter. A "no reaction" from someone with anaphylaxis is unambiguous, but coeliac harm is often asymptomatic — you can feel fine at the table and still have been exposed — so a coeliac "no reaction" counts for less per visit than an anaphylactic one. A reaction is an unambiguous negative and is never scaled down by severity.
What can a diner report as an outcome?
No reaction, a reaction, a near-miss ("something felt off"), or unsure. A near-miss lowers the signal as an early warning before anyone is harmed, and it lingers in the weighting longer than an ordinary visit.
How does recency affect confidence?
Allergy trust decays, because staff, chefs, suppliers and menus turn over. Recent reports weigh more heavily than old ones, evidence is labelled fresh, ageing or stale, and a venue with nothing fresh for around 18 months reads as "no recent evidence" rather than letting an old clean record pass as current.
What happens when there isn’t much evidence yet?
Below a handful of matching diners, the card shows "not enough reports from diners like you yet" instead of a number. feefrae also grades how much evidence exists per allergen — none, limited or strong — so a multi-allergen diner can see exactly where it is thin for their combination.
How do you check a venue’s own claims?
When a venue states something — a dedicated gluten-free fryer, separate preparation — diners who actually ate there say whether they saw it. feefrae shows the agreement or the gap between the claim and what diners observed, and grades how much evidence backs it. That grades the evidence, never whether the venue is safe.
What does the trend mean?
Diners trust a direction, not a snapshot. feefrae combines recent outcomes, freshness and any open change reports into a trend — improving, stable or declining. A declining trend or a fresh change report is the most decision-relevant thing the card can surface.
Can a venue pay to raise its confidence or remove a bad outcome?
No. The confidence signal is earned from diner evidence only. Venues can pay for clearly-labelled marketing placement and tools, and they can respond to a report — but money never buys the signal, an un-earned badge, or the removal of a negative outcome.
Does the confidence signal mean a venue is safe?
No. It is evidence to weigh, not a guarantee and not a verdict. feefrae never tells you what you can eat. Always confirm directly with the venue, and carry any medication you have been prescribed.
One thing we deliberately never do: publish a single “trust score”. A branded number gets read as a safety verdict, which is exactly what we refuse to issue. More on that in what we don’t do.